Steps and spikes in current-voltage
of oxide/microcrystallite-silicon/oxide
characteristics
diod&s
S. Y. Chou and A. E. Gordon
Department of Electrical Engineering, University of Minnesota, Minneapolis, Minnesota 5545.5
(Received 5 September 1991; accepted for publication 30 January 1992)
SiO,/microcrystallite-Si/SiO,
diodes with different contact window sizes were fabricated and
studied at room temperature, 77 K, and 4.2 K. Steps were observed in the current-voltage (I-V)
characteristics at all three temperatures. These steps would appear for a certain number of
measurements, depending upon measurement temperature, and then were replaced by a smooth
electrical breakdown I-V characteristic. Data analysis indicates that the steps in the 1-V
characteristics are due to local electrical breakdowns along the edge of metal contacts instead of
electron resonant tunneling through the structure. Surprisingly, however, in one diode, three
repeatable spikes, instead of steps, were observed at 4.2 K; this cannot be satisfactorily explained
in terms of electrical breakdown, and seems rather like electron resonant tunneling.
Excitement has been aroused by recent reports on
room-temperature observation of negative differential conductances (NDCs) in a structure consisting of a 5 nm
microcrystallite-silicon
sandwiched between two 2-nm
amorphous-S&C1 -,:H barriers,’ and in a structure consisting of an 8-nm microcrystallite-silicon
sandwiched between two 3-nm Si02 barriers2 These NDCs were attributed
to
electron
resonant
tunneling
through
microcrystallite (mc ) -silicon, leading to a possibility that
room-temperature quantum-effect devices could be realized in silicon-the
backbone material of the integrated
circuit industry.
In this letter, we present the study of diodes with a
structure of SiO,/mc-Si/SiO,, similar to the one in Ref. 2,
and report the observation of steps and spikes in the current-voltage (1-V) characteristics of these structures. We
will show that the steps can be explained by local oxide
breakdowns along the edge of the metal contact rather
than electron resonant tunneling. However, the spikes in
the I-V characteristics cannot be interpreted satisfactorily
using the breakdown model.
In fabrication of SiO,/mc-Si/SiO,
diodes, lOO-nm
thermal oxide was f&t grown, using a wet oxidation and
NZ annealing, on an n-type silicon substrate with a phosphorus doping concentration of 8X 1015 cm - 3. Then, the
SiOz on the backside of the wafer was etched away and the
backside was doped with phosphorous to form a n + layer
for a good back contact. Windows with different areas were
opened in Si02 on the front side using HF etching, followed by a depositing of 12-nm-thick polysilicon using an
electron-beam evaporation, with a substrate temperature of
-60 “C at a pressure of 2 X 10 - ’ Torr. The polysilicon
was then annealed and oxidized at 800 “C for 20 min in
N2/02 (with a 3:l ratio) at atmosphere pressure. Due to
fast diffusion of oxygen along the gram boundary, it is
believed that, after the oxidation, the mc-Si becomes embodied into SiO,. Comparing our fabrication process with
literature3 and extrapolating the anneal time, we estimate
the mc-Si to have an average thickness of -9 nm and each
oxide layer to have an average thickness of -4 nm.
Literature4 also indicates that the majority of these mc-Si
1827
should be in (111) orientation. Finally, the aluminum
(Al) gate and Al substrate contacts (after etching away
SiO, on the wafer backside) were deposited using electronbeam evaporation. The resulting structure is shown in
Fig. 1.
Seven sizes of devices were fabricated on the same substrate, all of them squares, with sides of 4, 5, 8, 12.5, 37.5,
65, and 87 pm. An HP-4145B semiconductor analyzer was
used in all measurements. The 77 and 4.2 K measurements
were carried out by immersing the bonded devices in liquid
nitrogen and liquid helium, respectively.
The I-V characteristics of a typical SiO,/mc-Si/SiO,
diode under forward bias conditions (i.e., the Al gate is
positive) are very similar to a Schottky diode with an ideality factor of 11, 50, and 4000 at room temperature, 77 K,
and 4.2 K, respectively.
Under a reverse bias and at room temperature, steps
were observed in the I-V characteristics during the first
measurement of the diodes (Fig. 2, curve a). In subsequent
measurements, the steps disappeared and a smooth breakdown diode 1-V characteristic resulted (Fig. 2, curve b) .
At 77 K, we found that the steps existed in three consecutive measurements before a breakdown I-V appeared and
that, in the repeated measurements, each step generally did
not occur at the same voltage as the previous measurement; they seemed rather random. At 4.2 K, the number of
measurements that the steps exist increased to about ten
before a breakdown diode I- V characteristic appeared. The
steps at all three temperatures looked similar.
To study the origins of the stairs in I-V characteristics,
icon dioxide
m&i
n-Silicon
aluminum
FIG.
1. Schematics of the cross section of an SiO,/
microcrystallite-Si/SiOz diode. The m&i has an average diameter of -9
nm and is sandwiched between -4 nm SiO,.
Appl. Phys. Lett. 60 (15), 13 April 1992
0003-6951/92/j
51827-03$03.00
@ 1992 American Institute of Physics
1827
Downloaded 06 Aug 2003 to 128.112.49.65. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
-1.5
-1.0
a
3
3 -0.5
& I5
4
VI
%
8
10
B
0.0
2
5
0.5
0
-10
-20
Gate bias 0’)
-30
-40
0
0
100
zoo
300
400
Pad Perimeter (pm)
FIG. 2. Room-temperature
current-voltage
curve of a SiOs/
microcrystallite-SiiSiO~ diode under a reverse bias. Curve a is the first
measurement and curve b is the second measurement.
we conducted the following data analysis. First, we made a
distribution of step height, and found that the distribution
is independent of device area and measurement temperature. A typical step-height distribution is shown in Fig. 3.
The average step height is - 180 PA. If a step was due to
tunneling through a silicon microcrystal, with a cross section of 9 by 9 nm’, then the current density would have
been 2x lo* A/cmZ, which is so large that the Si microcrystal would have been melted immediately. Second, we
plotted the current of the same diode in Fig. 2 (curve a) as
a function of l/V. It showed that at low reverse bias, the
diode current is logrithmically proportional to the inverse
of the gate voltage, indicating Fowler and Nordheim tunneling, and at high reverse bias, the current increases very
quickly with the gate voltage, indicating an impact ionization before the steps occur. Third, we plotted the number
of steps versus the perimeter size of the contact window
(Fig. 4) and found that they are proportional. Similar behavior was observed at 77 and 4.2 K. To determine the
error bars, 50 devices were measured for each perimeter
size. Fourth, we plotted the diode current after it had broken down (i.e., curve b in Fig. 2) at a given voltage versus
the perimeter of the contact window and found that it is
also proportional to the perimeter (Fig. 5). In other words,
the breakdown current per unit perimeter size versus voltage characteristics are in fact independent of the contactwindow size, indicating the breakdown current flow
through the edge of the gate metal contact. And finally, if
FIG. 4. Number of step vs perimeter of the contact window.
the steps were due to tunneling, their position should have
been very reproducible, instead of random.
All of the above five observations indicate that a step in
the I-V characteristic is not due to electron resonant tunneling but rather to a local electrical breakdown between
the Si substrate and a point somewhere along the edge of
the gate metal-contact pad. The sharp edge of a metal
contact has a smaller radius of curvature, and therefore a
higher electrical field. As a result, it is easier to electrically
break down. Some studies showed that sharp edges in the
metal contact can lower the oxide breakdown voltage by
one-half or two-thirds.s We believe that in our case the
breakdown starts at the weakest point of the structure
(most likely at the edge of the contact), progresses to the
next weakest point, and so on; each breakdown forms a
step in the diode I-V characteristics. At room temperature,
the breakdown current melts the material at the breakdown point due to Joule heating, and therefore, the breakdown is irreversible. This explains why the staircase could
be seen only once at the room-temperature ~I- V. Such irreversible melting breakdown (but not the staircase) has
been observed in metal-oxide-semiconductor systems.6 On
the other hand, at 77 and 4.2 K, the substrate is cooled by
surrounding liquid and current heating is much less. Because the cooling prevents immediate melting of materials,
the breakdown is initially an avalanche process and reversible, leading to repeatable staircase I-V characteristics. Obviously, the lower the temperature, the better the cooling
and the more repeatable the staircase 1-V characteristic.
This is exactly what has been observed in the experiments.
10
P 8
;2 6
g
z
4
2
0
100
200
300
400
Step Height in @A)
500
FIG. 3. Distribution of step height for an 87 by 87 pm’ device at room
temperature showing an average step height of 180 PA.
1828
Appl. Phys. Lett., Vol. 60, No. 15, 13 April 1992
Contact Perimeter Size (pm)
FIG. 5. The current in a diode already broken down at a fixed voltage vs
perimeter of the contact window.
S. Y. Chou and A. E. Gordon
1828
Downloaded 06 Aug 2003 to 128.112.49.65. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
-0.8
-0.6
3
-0.2
0
0
-10
-20
-30
-40
Gate bias (v)
FIG. 6. The I-V characteristics of a SiO,/microcrystallite-WSiO,
diode
of 4x4 pmZ area at 4.2 K showing three repeatable spikes. The voltage
increment during the measurements was 0.1 V.
i
To further prove that the steps in the diode I-V are due
to electrical breakdown and that the breakdown causes
material melting, we annealed the diodes that had broken
down in a furnace at 150 “C for 1 h in the air and found.
that, after annealing, the staircase I-V characteristics reappeared. This is what we expected since annealing can oxidize the molten materials at the breakdown regions, making them insulators again.
One surprise appeared in our experiment, however: at
4.2 K, one of the devices with a cross section of 4x4 pm”
fabricated on the same substrate showed three spikes in the
1-V characteristic instead of steps (Fig. 6 j. We made six
concurrent 1-V measurements on the device before thermally recycling it, and the three peaks occurred at the
same voltages as the previous measurements. The largest
peak-to-valley current ratio of the spikes is 11. These spikes
are difficult to understand in a local electrical-breakdown
model. First, in the local electrical breakdown the voltage
position for each breakdown should be random, but the
three spikes occurred at the same voltages in every measurement. Second, the spikes go up gradually and smoothly
instead of making large jumps as in the case of the staircase. Third, the current spike is about two orders of magnitude lower than the average step height ( 180 PA) for the
staircase observed in other devices. If the current is due to
tunneling through the Si microcrystal (9 x 9 nm) ,2 the current density for each spike is about 1 X lo6 A/cm2, a reasonable number for tunneling. And finally, in the local
electrical breakdown, the current step does not drop back
but, in the spikes, the current does come back after the
peak. We would say that the spikes in I-V characteristics
look more like resonant tunneling than local electrical
breakdown. The device was broken down when we thermally recycled it. More than 60 devices of the same cross
section were measured at 4.2 K; none of them showed the
spikes in the 1-V characteristics. This makes it difficult to
ascertain what caused the spikes; further investigation is in
progress.
In summary, we have fabricated Si02/mc-Si/Si02 diode structures with different contact-window sizes. We
have observed steps in the I-V characteristics at room temperature, 77 K, and 4.2 K. We have shown that these steps
in the I-V are due to local electrical breakdowns instead of
electron resonant tunneling. One device, however, showed
three repeatable spikes in the I-V characteristics at 4.2 K,
which resembles resonant tunneling rather than local electrical breakdown.
It is our pleasure to acknowledge M. I. Nathan for the
helpful discussions and R. Tsu for providing information
about his sample preparation. The work is partially supported by the Packard Foundation through a Packard Fellowship Award, by IBM through an IBM Faculty Development Award, and SRC under Contract No. 91-SJ-231.
‘E. Fortunato, R. Martins, I. Ferreira, M. Santos, A. Marcario, and L.
Guimaraes, J. Non-Cryst. Solids, 115, 120 (1989).
‘R. Tsu, Quiyi Ye, and E. H. Nicollian, SPIE, 1361, 232 (1990).
‘R. Tsu, J. Gonzalez-Hernandes, S. S. Chao, and D. Martin, Appl.
Phys. Lett. 48, 647 (1986).
4R. Tsu, J. Gonzalez-Hemandes, S. S. Chao, and D. Martin, Appl.
Phys. Lett. 46, 1089 (1985).
5N. Klien and 0. Nevanlinna, Solid-State Electron. 26, 883 (1983).
‘D. N. Chen and Y. C. Cheng, J. Appl. Phys. 61, 1592 (1987).
1829
Appl. Phys. Lett., Vol. 60, No. 15, 13 April 1992
S. Y. Chou and A. E. Gordon
1829
Downloaded 06 Aug 2003 to 128.112.49.65. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp